A Feasibility Study on The Implementation of Neural Network Classifiers for Open Stope Design

Abstract Assessing the stability of stopes is essential in open stope mine design as unstable hangingwalls and footwalls lead to sloughing, unplanned dilution, safety concerns compromising profitability mine. Over past few decades, numerous empirical tools have been developed dimension connection with its stability, using graph method. However, one principal limitations method objectively determine boundary zones, gain a clear probabilistic interpretation graph. To overcome this issue, paper aims explore feasibility artificial neural network (ANN) based classifiers for stopes. A database was compiled included dimensions, rock mass properties, conditions. The main parameters modified number (N’), conditions (stable, unstable, failed), hydraulic radius (HR). feed-forward (FFNN) classifier containing two hidden layers (110 neurons each) employed identify Overall, outcome analysis showed good agreement field data; most surfaces were correctly predicted an average accuracy 91%. This shows improvement over existing In addition, better results, associated probability occurrence stable, or caved determined shown iso-probability contour charts which compared proposed FFNN-based outperformed conventional terms prabablistic interpretation. It suggested that could be reliable tool can complement